Training a Supervised EdgeWise Model

17.3.9 Training a Supervised EdgeWise Model

You can train a SupervisedEdgeWiseModel on a graph as shown:

opg4j> model.fit(trainGraph)

model.fit(trainGraph);

model.fit(train_graph)

You can also add a validation step to the training. When training a model, the optimal number of training epochs is not known in advance and it is one of the key parameters that determines the model quality. Being able to monitor the training and validation losses helps you to identify a good value for the model parameters and gain visibility in the training process. The evaluation frequency can be specified in terms of epoch or step. To configure a validation step, create a GraphWiseValidationConfig and pass it to the model builder as shown:

opg4j> import oracle.pgx.config.mllib.EvaluationFrequencyScale;

opg4j> var validationConfig = analyst.graphWiseValidationConfigBuilder().
         setEvaluationFrequency(2).                                    // set the evaluation frequency (default: 1)
         setEvaluationFrequencyScale(EvaluationFrequencyScale.EPOCH).  // available options: EPOCH, STEP (default: EPOCH)
         build()

opg4j> var model = analyst.supervisedEdgeWiseModelBuilder().
         setVertexInputPropertyNames("vertex_features").
         setEdgeInputPropertyNames("edge_features").
         setEdgeTargetPropertyName("labels").
         setValidationConfig(validationConfig).  // configuring the validation to be executed every 2 epochs
         build()

import oracle.pgx.config.mllib.GraphWiseValidationConfig;
import oracle.pgx.config.mllib.EvaluationFrequencyScale;

GraphWiseValidationConfig validationConfig = analyst.graphWiseValidationConfigBuilder()
    .setEvaluationFrequency(2)                                    // set the evaluation frequency (default: 1)
    .setEvaluationFrequencyScale(EvaluationFrequencyScale.EPOCH)  // available options: EPOCH, STEP (default: EPOCH)
    .build();

SupervisedEdgeWiseModel model = analyst.supervisedEdgeWiseModelBuilder()
    .setVertexInputPropertyNames("vertex_features")
    .setEdgeInputPropertyNames("edge_features")
    .setEdgeTargetPropertyName("labels")
    .setValidationConfig(validationConfig)  // configuring the validation to be executed every 2 epochs
    .build();

validation_config = analyst.graphwise_validation_config(
    evaluation_frequency=2,              # set the evaluation frequency (default: 1)
    evaluation_frequency_scale="epoch",  # available options: "epoch", "step" (default: "epoch")
)

params = dict(edge_target_property_name="labels",
              vertex_input_property_names=["vertex_features"],
              edge_input_property_names=["edge_features"],
              validation_config=validation_config,  # configuring the validation to be executed every 2 epochs
              seed=17)

model = analyst.supervised_edgewise_builder(**params)

After configuring a validation step, you can then pass a graph for validation to the fit method together with the graph for training:

opg4j> model.fit(trainGraph, valGraph)

model.fit(trainGraph,valGraph);

model.fit(train_graph,valGraph)

Parent topic: Using the Supervised EdgeWise Algorithm (Edge Embeddings and Classification)